On 10/8/21 09:20, Thomas Huth wrote:
On 07/10/2021 12.44, Janosch Frank wrote:
On 10/7/21 11:44, Thomas Huth wrote:
On 07/10/2021 10.50, Janosch Frank wrote:
Let's just define all of the needed things so we can link libcflat.
A significant portion of the lib won't work, like printing and
allocation but we can still use things like memset() which already
improves our lives significantly.
Signed-off-by: Janosch Frank <frankja@xxxxxxxxxxxxx>
---
s390x/snippets/c/cstart.S | 14 ++++++++++++++
1 file changed, 14 insertions(+)
diff --git a/s390x/snippets/c/cstart.S b/s390x/snippets/c/cstart.S
index 031a6b83..2d397669 100644
--- a/s390x/snippets/c/cstart.S
+++ b/s390x/snippets/c/cstart.S
@@ -20,6 +20,20 @@ start:
lghi %r15, stackptr
sam64
brasl %r14, main
+/*
+ * Defining things that the linker needs to link in libcflat and make
+ * them result in sigp stop if called.
+ */
+.globl sie_exit
+.globl sie_entry
+.globl smp_cpu_setup_state
+.globl ipl_args
+.globl auxinfo
+sie_exit:
+sie_entry:
+smp_cpu_setup_state:
+ipl_args:
+auxinfo:
I think this likely could be done in a somewhat nicer way, e.g. by moving
Definitely, as I said, it's a simple fix
Alternatively, something like this might work, too:
Seems like it works for the two tests that I checked.
Would you mind sending a proper patch?
I'd like to send out a pull today or on Monday.
diff --git a/s390x/Makefile b/s390x/Makefile
--- a/s390x/Makefile
+++ b/s390x/Makefile
@@ -80,7 +80,7 @@ asmlib = $(TEST_DIR)/cstart64.o $(TEST_DIR)/cpu.o
FLATLIBS = $(libcflat)
SNIPPET_DIR = $(TEST_DIR)/snippets
-snippet_asmlib = $(SNIPPET_DIR)/c/cstart.o
+snippet_asmlib = $(SNIPPET_DIR)/c/cstart.o lib/auxinfo.o
# perquisites (=guests) for the snippet hosts.
# $(TEST_DIR)/<snippet-host>.elf: snippets = $(SNIPPET_DIR)/<c/asm>/<snippet>.gbin
diff --git a/s390x/snippets/c/cstart.S b/s390x/snippets/c/cstart.S
--- a/s390x/snippets/c/cstart.S
+++ b/s390x/snippets/c/cstart.S
@@ -21,5 +21,9 @@ start:
sam64
brasl %r14, main
/* For now let's only use cpu 0 in snippets so this will always work. */
+.global puts
+.global exit
+puts:
+exit:
xgr %r0, %r0
sigp %r2, %r0, SIGP_STOP
I think that's more clear this way, since we're fencing the
functions that caused the dependencies to the other functions
from your patch. What do you think?
Thomas
